Fault Detection Of Reciprocating Piston Diaphragm Pump Based On Multivariate Statistics

Along with the computer, mechanical design, communication network, such as materials science and technology rapid development, has prompted the modern industrial upgrading of production technology, control system of industrial automation, informationization level enhances unceasingly. Industrial production system is becoming more and more efficient and intelligent, the complexity of the corresponding is also more and more high. Slurry pipeline transport system as well. Slurry pipeline transportation costs have a bottom, high transport efficiency and environmental pollution and so on. Making the slurry pipeline to become the fifth largest transport modes road, sea, rail, air later. Pipeline can be long-distance, all-terrain transport can transport highly corrosive, strong abrasive to various solid-liquid transport minerals. It is a distinct advantage in some underdeveloped traffic, terrain, areas of high environmental requirements. In recent years, the rapid development of slurry pipeline transportation, and in the slurry pipeline in reciprocating membrane pumps are the main source of power, it’s the normal operation of the entire slurry pipeline transportation plays a key role. The safe operation of reciprocating membrane pump monitoring and fault detection have great significance.Through analysis and study in the analysis of the structure, operation principle and characteristics of fault signals in slurry pipeline transportation of reciprocating membrane pumps. Found that the traditional method based on signal processing in the presence of too much focus on the extraction and processing of a single vibration signal characteristics. Lack of reciprocating machinery overall grasp. Therefore, the use of multivariate statistical methods in this paper to study reciprocating membrane pump fault detection. In this paper, main work is as follows:1) For reciprocating membrane pump structure, operation principle and the SCADA system is studied and analyzed. Analysis of the failure of the signal characteristics and the main problem for the mainstream method of reciprocating membrane pumps troubles. For the following study of the made a certain base2) Reciprocating membrane pumps for wear in the early stages of development is difficult to fault by changing the observed variables were found, but the observation was high and strong nonlinear coupling between the variable output. Method based on kernel principal component analysis (KPCA) reciprocating membrane pumps wear fault detection model. First, historical monitoring data using reciprocating membrane pumps running establishment KPCA fault detection model, and use the model to detect the fault features. Then get around the contribution of the original variables and analyzed to locate the position of the occurrence of wear. Experimental results show that the method of reciprocating membrane pumps wear fault detection is effective.3) In the actual pipeline monitoring signals during reciprocating membrane pumps are often non-linear, non-Gaussian distribution, high-dimensional data, the failure occurred at the beginning, the characteristic signal is very small, it is difficult to be detected. And traditional multivariate statistical method is not effective to solve this problem. Joined the local learning and independent thinking subspace, using a fault detection method based on independent component subspace of a local modeling. Firstly, local LSSVR model output to predict various monitoring instruments, and with the formation of the observed output residuals compared using independent subspace then the sequence space into several different subspace, and then use ICA-PCA respectively establish fault detection model, and finally according to the situation of each sub-space statistic to detect. Simulation results show that the method of small fault has a certain fault detection.This paper studies reciprocating membrane pump fault detection and diagnosis through the use of multivariate statistical methods, to provide a theoretical basis for the realization of fault detection and diagnosis in the pipeline.